Main

SARS-CoV, SARS-CoV-2 and Middle East respiratory syndrome coronavirus (MERS-CoV) emerged from transmission events in which humans were infected with bat or camel coronaviruses5,6,7,8. Betacoronaviruses that circulate in civets, bats and Malayan pangolins are genetically similar to SARS-CoV and SARS-CoV-2, and use human ACE2 as a receptor5,6,9,10,11; these SARS-related coronaviruses have the potential to be transmitted to humans12. Cross-neutralizing antibodies (cross-nAbs) that are capable of neutralizing multiple betacoronaviruses, and of preventing or treating betacoronavirus infection, have been isolated from humans infected with SARS-CoV13,14,15,16,17,18,19,20,21,22,23,24, which provides proof-of-concept for the development of vaccines against members of the Sarbecovirus subgenus25.

In mice, the induction of cross-nAbs though vaccination has previously been reported for coronavirus pseudoviruses26,27. However, it is unknown whether the vaccination of primates with spike protein can elicit cross-nAbs against SARS-CoV, bat betacoronaviruses or SARS-CoV-2 escape viruses. One target of cross-nAbs is the receptor-binding domain (RBD) of the spike protein14,24,25: for example, the antibody DH1047 is a cross-nAb that targets the RBD and cross-neutralizes SARS-CoV, SARS-CoV-2 and bat coronaviruses15. RBD immunogenicity can be augmented by arraying multiple copies of the domain on nanoparticles, which mimics virus-like particles26,27,28,29. We therefore designed a vaccine that comprised a 24-mer SARS-CoV-2 RBD nanoparticle conjugated to a ferritin scaffold. We constructed the RBD nanoparticle by expressing recombinant SARS-CoV-2 RBD with a C-terminal sortase A donor sequence, and by expressing a 24-subunit, self-assembling Helicobacter pylori ferritin with an N-terminal sortase A acceptor sequence30. We then conjugated the RBD to ferritin nanoparticles using a sortase A reaction30 (Fig. 1a, Extended Data Fig. 1). We used analytical size-exclusion chromatography and negative-stain electron microscopy to confirm that the RBD was conjugated to the surface of the ferritin nanoparticle (Fig. 1a, Extended Data Fig. 1b, c). The RBD sortase-A-conjugated nanoparticle (hereafter, RBD–scNP) bound to human ACE2 (the receptor for SARS-CoV-2) and to potent neutralizing SARS-CoV-2-specific RBD antibodies DH1041, DH1042, DH1043, DH1044 and DH104515 (Fig. 1b). The cross-nAb DH1047 also bound to the RBD–scNP (Fig. 1b). The RBD–scNP lacked binding to SARS-CoV-2 spike antibodies that bound outside of the RBD (Fig. 1b).

Fig. 1: RBD–scNP elicits extremely high titres of SARS-CoV-2-pseudovirus neutralizing antibodies.
figure 1

a, SARS-CoV-2 RBD (blue and red)–H. pylori ferritin (grey) nanoparticle sortase (SrtA) conjugation. Model and two-dimensional class average of negative-stain electron microscopy (EM) of the resultant RBD–scNP are shown. b, Biolayer interferometry SARS-CoV-2 antibody and ACE2-receptor binding to RBD nanoparticles. nAbs, neutralizing antibodies; non-nAbs, non-neutralizing antibodies; non-nAbs IE, infection-enhancing non-neutralizing antibody; NTD, N-terminal domain. Symbols represent values from three independent experiments; data are mean + s.e.m. c, Design of the cynomolgus macaque immunogenicity and challenge study. d, Macaque serum IgG binding titre as area-under-the-curve (AUC) of the log10-transformed curve (log(AUC)) to recombinant SARS-CoV-2 S-2P, RBD, NTD and fusion peptide. Group mean ± s.e.m. are shown in d, e (n = 5 macaques). e, Plasma antibody blocking of SARS-CoV-2 S-2P binding to ACE2 and RBD neutralizing antibody DH1041. f, g, Dose-dependent serum neutralization of SARS-CoV-2 D614G pseudovirus infection of ACE2-expressing 293T cells (f) and neutralization ID50 and ID80 titres (g). Serum was examined after two immunizations. The mean value of duplicates is shown in f. h, SARS-CoV-2 D614G pseudovirus serum neutralization titre over time for individual macaques. i, Serum neutralization ID50 titres from macaques immunized twice with RBD–scNP (blue) or S-2P mRNA-LNP (burgundy). **P = 0.0079, two-tailed exact Wilcoxon test. n = 5 macaques. j, Serum neutralization titres for macaques immunized twice with RBD–scNP (blue) (n = 5 macaques) or humans with asymptomatic infection (n = 34 individuals) (A), symptomatic infection (n = 71 individuals) (S) or who were hospitalized (n = 60 individuals) (H). **P < 0.01, two-tailed Wilcoxon test. Horizontal bars are the group geometric mean in i, j. Pre-vaccination serum or neutralizing antibody spiked serum were used as controls in fhd, e, h, Arrows, time of immunization.

Source data

We immunized five cynomolgus macaques three times intramuscularly, four weeks apart, with 100 μg of RBD–scNP adjuvanted with 5 μg of the TLR7 and TLR8 agonist 3M-052 absorbed to 500 μg of alum31 (Fig. 1c, Extended Data Fig. 1d, e). Immunizations were well-tolerated in macaques (Extended Data Fig. 2). Immunization with RBD–scNP adjuvanted with 3M-052 and alum elicited binding IgG against both the SARS-CoV-2 RBD and the spike ectodomain stabilized by the introduction of two prolines (S-2P) (Fig. 1d), but immunization with 3M-052 and alum alone did not (Extended Data Fig. 3a, b). Boosting once maximally increased SARS-CoV-2 binding IgG titres (Fig. 1d). ACE2 competitive-binding assays demonstrated the presence of ACE2-binding-site antibodies in the serum of vaccinated macaques (Fig. 1e). Similarly, plasma antibodies blocked the binding of the ACE2-binding-site-focused, RBD-neutralizing antibody DH1041 (Fig. 1e). We assessed the vaccine induction of neutralizing antibodies against a SARS-CoV-2 pseudovirus with an aspartic acid-to-glycine substitution at position 614 (D614G)32. Two RBD–scNP immunizations induced potent serum neutralizing antibodies, with ID50 neutralization titres that ranged from 21,292 to 162,603 (Fig. 1f, g). We compared these neutralizing antibody titres to those elicited by cynomolgus macaques that were immunized twice with 50 μg of nucleoside-modified mRNA encapsulated in lipid nanoparticles that encodes transmembrane S-2P (S-2P mRNA-LNP). S-2P mRNA-LNP is analogous to COVID-19 vaccines that have been authorized for emergency use (Extended Data Fig. 1f). The serum neutralization titres against SARS-CoV-2 D614G pseudovirus that were elicited by RBD–scNP immunization were significantly higher than titres that were elicited by two immunizations with S-2P mRNA-LNP33,34 (group geometric mean ID50 of 47,216 and 6,469, respectively; P = 0.0079 exact Wilcoxon test, n = 5 macaques for each vaccine) (Fig. 1i). When compared to natural human infection, RBD–scNP vaccination elicited higher ID50 neutralization titres (Fig. 1j). RBD–scNP adjuvanted with 3M-052 and alum therefore elicits significantly higher neutralizing titres in macaques, compared to current vaccine platforms or natural human infection (Fig. 1i, j).

The SARS-CoV-2 variant B.1.1.7 is spreading globally, and has previously been suggested to have higher infectivity than the Wuhan-Hu-1 strain35,36. B.1.351-lineage viruses are widespread in the Republic of South Africa36,37,38 and—along with the P.1 variant—are of concern owing to their neutralization-resistant phenotype, which is mediated by K417N, E484K and N501Y substitutions in the RBD39. Each of these mutations is distal to the cross-nAb DH1047 binding site (owing to its long HCDR3 domain that is used to contact the RBD); however, the E484K substitution is within the binding site of the RBD neutralizing antibody DH104115 (Fig. 2a, b). DH1041 binding to the SARS-CoV-2 RBD was therefore knocked out by the E484K substitution, but DH1047 binding to the RBD was unaffected by the K417N, E484K or N501Y substitutions (Fig. 2c, d, Extended Data Fig. 3).

Fig. 2: RBD–scNP elicits a higher titre of neutralizing antibodies against more transmissible or neutralization-resistant SARS-CoV-2 variants than does S-2P mRNA-LNP.
figure 2

a, b, The location of K417, E484 and N501 (spheres), which are mutated in the B.1.351 variant, are shown in the cryo-electron microscopy structures of the RBD neutralizing antibodies DH1041 (red) (a) and DH1047 (magenta) (b) bound to the RBD (grey) of spike trimers (Protein Data Bank codes (PDB) 7LAA and 7LD1)15. c, ACE2 receptor, DH1041 and DH1047 enzyme-linked immunosorbent assay (ELISA) binding titre as log(AUC), for wild-type (WT) and mutant SARS-CoV-2 spike RBD monomers. d, Serum neutralization ID50 (left) and ID80 (right) titres for SARS-CoV-2 D614G and SARS-CoV-2 B.1.1.7 pseudoviruses from immunized macaques, in ACE2-expressing 293 cells. Symbols represent individual macaques; horizontal bars are group means. **P = 0.0079, two-tailed exact Wilcoxon test. n = 5 macaques. e, Fold decrease in neutralization potency in neutralization of SARS-CoV-2 B.1.1.7 pseudovirus relative to SARS-CoV-2 D614G pseudovirus. Fold change is shown for RBD–scNP-immunized and S-2P mRNA-LNP-immunized macaques on the basis of the ID50 and ID80 titres. Horizontal bars are the group mean. f, Vaccinated macaque serum neutralization ID50 (left) and ID80 titres (right) against SARS-CoV-2 WA-1 and B.1.351 pseudoviruses, in ACE2-expressing 293 cells. Symbols and horizontal bars are shown the same as in d. *P = 0.0159; **P = 0.0079, two-tailed exact Wilcoxon test. n = 5 macaques. g, Fold decrease in neutralization potency in neutralization of SARS-CoV-2 B.1.351 pseudovirus relative to SARS-CoV-2 WA-1 pseudovirus. Fold change is shown as in e. h, Vaccine-induced ID50 (left) and ID80 (right) neutralization titres of infection of ACE2- and TMPRSS2-expressing 293 cells by SARS-CoV-2 WA-1 or P.1 pseudovirus. Symbols and horizontal bars are as in d. *P = 0.0159; **P = 0.0079, two-tailed exact Wilcoxon test. n = 5 macaques. i, Fold decrease in neutralization potency (as shown in e) for SARS-CoV-2 P.1 pseudovirus relative to SARS-CoV-2 WA-1 pseudovirus.

Source data

We determined whether immunization with RBD–scNP or S-2P mRNA-LNP elicited neutralizing antibodies against these particular SARS-CoV-2 variants. Serum from macaques vaccinated with RBD–scNP potently neutralized a pseudovirus bearing the D614G spike or the B.1.1.7 spike (Fig. 2d, e). Similarly, S-2P mRNA-LNP immunization elicited equivalent titres of neutralizing antibodies against the B.1.1.7 and D614G variants, although titres were lower than with RBD–scNP immunization (Fig. 2d, e). Serum from macaque immunized with RBD–scNP or S-2P mRNA-LNP neutralized SARS-CoV-2 WA-1, B.1.351 and P.1 pseudoviruses, with 80% inhibitory reciprocal serum dilution (ID80) titres being more potent for the RBD–scNP group (Fig. 2f–i). On average, the neutralization titres in the RBD–scNP group decreased by threefold against the B.1.351 or P.1 variants, whereas those of the S-2P mRNA-LNP group decreased by sixfold for B.1.351 and tenfold for P.1 (on the basis of ID50 titres) (Fig. 2g, i). Additionally, we observed that plasma IgG binding to SARS-CoV-2 spike after RBD–scNP and S-2P mRNA-LNP immunization was unaffected by mutations that have been observed in SARS-CoV-2 from Danish minks, as well as in the B.1.351, P.1 and B.1.1.7 strains36,37,40 (Extended Data Fig. 3). In summary, both of the vaccines we tested elicited neutralizing antibodies that were unaffected by the mutations in the B.1.1.7 strain. However, neutralizing antibodies elicited by RBD–scNP more potently neutralized the B.1.351 and P.1 virus strains than did neutralizing antibodies elicited with S-2P mRNA-LNP immunization.

SARS-related coronaviruses that circulate in humans and animals remain a threat for future outbreaks12,41,42. We therefore examined the neutralization of SARS-CoV, the SARS-related group-2b bat coronavirus WIV-1 and the SARS-related bat coronavirus SHC014 by immune sera from macaques vaccinated with the RBD–scNP, S-2P mRNA-LNP or mRNA-LNP encoding monomeric RBD (RBD mRNA-LNP)6,9,41,42 (Extended Data Fig. 1e–g). After two immunizations, RBD–scNP, S-2P mRNA-LNP and RBD mRNA-LNP elicited neutralizing antibodies against SARS-CoV, WIV-1 and SHC014 (Fig. 3a, Extended Data Fig. 4). Neutralization was more potent for replication-competent SARS-CoV-2 virus compared to these three SARS-related viruses (Fig. 3a, Extended Data Fig. 4), and neutralization titres varied up to fourfold within the RBD–scNP group (Extended Data Fig. 4). Overall, RBD–scNP immunization elicited the highest neutralization titres (Fig. 3a, Extended Data Fig. 4). Modest increases in neutralization potency were gained by boosting a third time with RBD–scNP (Fig. 3b). Moreover, RBD–scNP immunization elicited cross-reactive plasma IgG against the spike proteins of SARS-CoV-2 and SARS-CoV, as well as those of the bat coronaviruses RaTG13 and SHC014 and the pangolin coronavirus GXP4L (Fig. 3c, Extended Data Fig. 5a, c). Binding antibody titres were high for these spike proteins, even in instances in which neutralization titres were low; this suggests that non-neutralizing antibodies contributed to binding titres. RBD–scNP-immune plasma IgG did not bind the spike protein from the four endemic human coronaviruses or MERS-CoV (Extended Data Fig. 5a, c). The lack of binding by plasma IgG to spike ectodomains of these latter five coronaviruses is consistent with RBD-sequence divergence among group 1, 2a, 2b and 2c coronaviruses (Fig. 3f, Extended Data Figs. 6, 7). The SARS-CoV-2 spike induced cross-nAbs against several group-2b SARS-related betacoronaviruses, with the highest titres being induced by RBD–scNP.

Fig. 3: Serum cross-neutralization of infections with SARS-related betacoronaviruses induced by RBD–scNP.
figure 3

a, Serum neutralization ID50 titres from macaques immunized twice with RBD–scNP, S-2P mRNA-LNP or RBD mRNA-LNP for SARS-CoV, SARS-CoV-2 and SARS-related bat coronaviruses (WIV-1 and SHC014) infections. Symbols indicate individual macaques; black bars show the group geometric mean. Two-tailed exact Wilcoxon test, n = 5 or 8 macaques. b, Serum cross-neutralization ID50 titres before (grey), or after two (light blue) or three (blue), RBD–scNP immunizations. Bars represent the group geometric mean. c, Binding titre (log(AUC)), based on ELISA for the spike protein of human, bat (RaTG13 and SHC014) and pangolin (GXP4L) SARS-related betacoronaviruses, for plasma IgG from macaques immunized twice with RBD–scNP. ECD, ectodomain; S, spike. d, Structural comparison of the epitopes of SARS-CoV-2-specific neutralizing RBD antibody (DH1041) (red) (PDB 7LAA) and cross-neutralizing RBD antibody (DH1047) (magenta) (PDB 7LD1). Left, cartoon view of spike (green), RBD (grey) and receptor binding motif (RBM) (blue). Right, overlay of the RBDs of the two complexes from their respective cryo-electron microscopy structures. e, RBD coloured by conservation within group-2b betacoronaviruses. DH1047 epitope is shown in magenta outline. f, RBD sequence conservation. Heat map displaying pairwise amino acid sequence similarity for 57 representative betacoronaviruses. g, h, Plasma or serum antibody blocking of S-2P binding to ACE2 (grey) and DH1047 (navy blue). g, Plasma antibody inhibition of spike binding. SARS-CoV-2 S-2P (left) or SHC014 S-2P (right) blocking kinetics by serum from macaques immunized twice with RBD–scNP. Group mean ± s.e.m. is shown. n = 5 macaques. Arrows, time of immunization. h, Blocking of SARS-CoV-2 S-2P binding to ACE2 (left) or DH1047 (right). Blocking activity by serum from macaques immunized twice with RBD–scNP or S-2P mRNA-LNP, and humans immunized twice with Pfizer BNT162b2 (BNT162b) or naturally infected with SARS-CoV-2 (convalescent). Each symbol represents an individual and filled bars indicate the group mean in h. Positivity threshold (dashed line) is greater than 20% in h.

Source data

Immune sera from RBD–scNP-immunized macaques exhibited a cross-neutralizing profile similar to that of the cross-nAb DH1047. DH1047 bound with <0.02 nM affinity to monomeric SARS-CoV-2 RBD (Extended Data Fig. 5b), and bound the RBD–scNP (Fig. 1b). The cross-reactive DH1047 epitope is adjacent to the N terminus of the ACE2-binding site, which distinguishes the DH1047 antibody from dominant ACE2-binding- site-focused neutralizing antibodies such as DH104115 (Fig. 3d), and the DH1047 epitope has high group-2b sequence conservation (Fig. 3e). Overall RBD sequences within betacoronavirus groups are more conserved than sequences from different groups (Fig. 3f, Extended Data Figs. 6, 7). We determined the presence of DH1047-like antibodies using DH1047 blocking assays. Plasma from all RBD–scNP-immunized macaques blocked the binding of ACE2 and DH1047 to SARS-CoV-2 S-2P (Fig. 3g, Extended Data Fig. 5d). The DH1047-blocking antibodies were cross-reactive, as they also potently blocked DH1047 binding to the S-2P of SHC014 (Fig. 3g). RBD–scNP immunization of macaques elicited higher magnitudes of DH1047-blocking antibodies than did S-2P mRNA-LNP; these magnitudes were also higher than those elicited by immunization of humans with the Pfizer BNT162b2 vaccine or SARS-CoV-2 infection in humans (Fig. 3h, Extended Data Fig. 5d). ACE2 blocking was high in all groups (Fig. 3h). Whereas 5 out of 5 RBD–scNP-vaccinated macaques exhibited potent DH1047 serum blocking activity, 3 out of 4 immunized humans and 9 out of 22 humans who had recovered from COVID-19 had detectable serum DH1047 blocking activity (Fig. 3h). The DH1047-like antibody response was therefore weak and subdominant in naturally infected or immunized humans and S-2P mRNA-LNP-immunized macaques, but was a dominant macaque antibody response to RBD–scNP vaccination.

To determine vaccine protection against coronavirus infection, we challenged RBD–scNP-vaccinated or S-2P mRNA-LNP-primed and RBD–scNP-boosted macaques with 105 plaque-forming units of SARS-CoV-2 virus via the intratracheal and intranasal routes, after their last boost (Fig. 4a). Neutralizing antibodies were detectable in all macaques at two weeks after the final immunization (Fig. 3b, Extended Data Fig. 4b, c). We collected bronchoalveolar lavage (BAL) fluid two days after the challenge (Fig. 4a), and detected infectious SARS-CoV-2 in BAL fluid from 5 out of 6 unimmunized macaques, but none of the RBD–scNP or S-2P mRNA-LNP and RBD–scNP-immunized macaques (Fig. 4b). We quantified SARS-CoV-2 replication as the number of copies of envelope (E) and nucleocapsid (N) subgenomic RNA (sgRNA) in fluid from nasal swabs and BAL at two and four days after challenge (Fig. 4a). At two days after challenge, unimmunized macaques had an average of 1.3 × 105 and 1.2 × 104 copies per ml of E sgRNA in the nasal swab and BAL fluids, respectively (Fig. 4c, d). By contrast, all of the RBD–scNP-vaccinated macaques, and 4 out of 5 macaques vaccinated with S-2P mRNA-LNP and RBD–scNP, had undetectable levels of E sgRNA in the upper and lower respiratory tract (Fig. 4c, d). We sampled macaques again two days later, and found no detectable E sgRNA in the BAL or nasal swab samples from any of the vaccinated macaques (Fig. 4b, c). Similarly, 4 out of 5 RBD–scNP-vaccinated macaques had undetectable N sgRNA in BAL and nasal swab fluid; the exception had 234 copies per ml of N sgRNA, which was detected on day 2 in nasal swab fluid (Fig. 4e, f). Virus replication was undetectable in this macaque by the fourth day after challenge (Fig. 4e). Additionally, all but one of the macaques immunized with S-2P mRNA-LNP and RBD–scNP had undetectable N sgRNA in BAL or nasal swab samples (Fig. 4e,f). Moreover, we did not detect SARS-CoV-2 nucleocapsid antigen in the lung tissue of any of the vaccinated macaques, but detected this antigen in all of the control macaques (Fig. 4g, Extended Data Fig. 8). Haematoxylin and eosin staining of lung tissue showed a reduction in inflammation in immunized macaques compared to control macaques (Extended Data Fig. 8, Extended Data Table 1).

Fig. 4: RBD–scNP vaccination alone or as a boost can prevent virus replication in the upper and lower respiratory tract after intranasal and intratracheal SARS-CoV-2 challenge in macaques.
figure 4

a, Study design for intranasal and intratracheal challenge of macaques with SARS-CoV-2. Blue and burgundy arrows indicate the time points for RBD–scNP and S-2P mRNA-LNP immunizations, respectively. b, Infectious virus in macaque BAL fluid two days after challenge. LOD, limit of detection. cf, Quantification of viral E gene (c, d) or N gene (e, f) sgRNA in unimmunized (grey) and RBD–scNP-immunized (blue) macaques, and in macaques with a S-2P mRNA-LNP prime and RBD–scNP boost (burgundy). sgRNA in nasal swabs (c, e) and BAL fluid (d, f) was quantified two (left panels) and four (right panels) days after challenge. LOD for the assay is 150 copies per ml. Symbols and bars represent individual macaques and group mean, respectively. g Nucleocapsid immunohistochemistry of lung tissue sections at seven days after challenge. Left, representative image from 1 macaque from each group of 5 macaques is shown. Red arrows indicate site of antigen positivity. All images are shown at 10× magnification. Scale bars, 100 μm. Right, quantification of lung viral antigen positivity. In each panel, symbols represent individual macaques (group mean is shown as a black horizontal bar).

Source data

Finally, when possible, we examined mucosal immunity to SARS-CoV-2 both before and after SARS-CoV-2 challenge (Extended Data Fig. 9). IgG from concentrated BAL bound to spike and blocked ACE2, DH1041 and DH1047 binding to spike (Extended Data Fig. 9b–d). Each response was higher in the BAL from macaques immunized three times with RBD–scNP, compared to macaques immunized two times with S-2P mRNA-LNP and boosted once with RBD–scNP (although the BAL was collected from each group at different time points). Unconcentrated nasal wash samples from macaques immunized with either RBD–scNP or a S-2P mRNA-LNP prime and RBD–scNP boost showed similar low levels of spike-binding IgG after challenge (Extended Data Fig. 9e). Nonetheless, RBD–scNP immunization elicited RBD-specific mucosal antibodies.

As three coronavirus epidemics have occurred in the past 20 years, there is a need to develop vaccines that are effective against all coronaviruses before the next pandemic25. The epitopes of betacoronavirus cross-nAbs (such as DH1047) provide clear targets for vaccines that aim to protect against multiple coronaviruses13,14,15,43. We have shown that immunization of macaques with RBD–scNP adjuvanted with 3M-052 and—to a lesser extent—S-2P mRNA-LNP induces cross-nAbs against multiple SARS-related human and bat betacoronaviruses. These results demonstrate that SARS-CoV-2 vaccination with either the RBD–scNP or S-2P mRNA-LNP vaccines (the latter of which are similar to vaccines that have already been authorized for use in humans) will probably elicit cross-nAbs, and have the potential to prevent future spillovers of group-2b betacoronaviruses from bats to humans12,26.

The emergence of neutralization-resistant and highly infectious variants of SARS-CoV-2 continues to be a concern for vaccine efficacy. The RBD–scNP and S-2P mRNA-LNP immunizations elicited SARS-CoV-2 neutralizing antibodies against the D614G, B.1.1.7, P.1 and B.1.351 strains of SARS-CoV-2. The neutralizing antibodies elicited by RBD–scNP and S-2P mRNA-LNP were of different specificities, as RBD–scNP-induced neutralizing antibodies showed a smaller reduction in neutralization potency across the different variants compared to S-2P mRNA-LNP immune sera. Our results are consistent with the previous demonstration that current COVID-19 vaccines have reduced efficacy against the B.1.351 SARS-CoV-2 variant39,44,45,46,47,48.

The RBD–scNP vaccine is a promising platform for the development of vaccines that target multiple coronaviruses for the following reasons. The RBD–scNP vaccine induced apparent sterilizing immunity in the upper respiratory tract, which has not been routinely achieved with SARS-CoV-2 vaccination in macaques49,50. Additionally, the high neutralization titres achieved by RBD–scNP vaccination bode well for an extended duration of protection. Despite the induction of high levels of antibody, we observed no evidence of increased immunopathology, inflammatory cytokines or virus replication indicative of vaccine-elicited antibody-dependent enhancement. This lack of in vivo infection enhancement is consistent with previous studies using SARS-CoV-2 monoclonal antibodies15. 3M-052 adsorbed to alum is in clinical testing (NCT04177355), which generates a potential translational pathway for RBD–scNP adjuvanted with 3M-052. The RBD–scNP vaccine represents a platform for inducing high titres of protective antibodies against current human coronaviruses and their variants, and for producing vaccines that could prevent, rapidly temper or extinguish the next spillover of a coronavirus into humans.

Methods

The experiments were not randomized, and only the pathologists were blinded to group treatments during experiments and outcome assessment.

Animals, immunizations, and human samples

Rhesus and cynomolgus macaques were housed and treated in Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC)-accredited institutions. The study protocol and all veterinarian procedures were approved by the Bioqual IACUC per a memorandum of understanding with the Duke IACUC, and were performed based on standard operating procedures. Macaques studied were housed and maintained in an AAALAC-accredited institution in accordance with the principles of the NIH. All studies were carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the NIH in BIOQUAL. BIOQUAL is fully accredited by AAALAC and through OLAW, assurance number A-3086. All physical procedures associated with this work were done under anaesthesia to minimize pain and distress, in accordance with the recommendations of the Weatherall report ‘The use of non-human primates in research.’ Teklad 5038 primate diet was provided once daily by macaque size and weight. The diet was supplemented with fresh fruit and vegetables. Fresh water was given ad libitum. All macaques were maintained in accordance with the Guide for the Care and Use of Laboratory Animals. mRNA-LNP was prepared as previously stated51,52. Rhesus macaques (n = 8 macaques) were immunized intramuscularly twice with 50 μg of RBD mRNA-LNP. Cynomolgus macaques (n = 5 macaques) were immunized twice with 50 μg of S-2P mRNA-LNP (encoding the transmembrane spike protein stabilized with K986P and V987P mutations) and boosted once with 100 μg of RBD–scNP adjuvanted with 5 μg of 3M-052 aqueous formulation admixed with 500 μg of alum in PBS. An additional group of cynomolgus macaques (n = 5 macaques) were immunized in the right and left quadriceps with 100 μg of RBD–scNP adjuvanted with 5 μg of 3M-052 aqueous formulation admixed with 500 μg of Alum in PBS31. The mixture for immunization consisted of 250 μl of RBD–scNP mixed with 250 μl of 0.02 mg ml−1 3M-052 and 2 mg ml−1 alum. Group sizes were selected such that statistical significance could be reached with between group nonparametric statistical comparisons. No other statistical methods were used to predetermine sample size. Macaques were on average 8 or 9 years old and ranged from 2.75 to 8 kg in body weight. Male and female macaques per group were balanced when availability permitted. Studies were performed unblinded. Macaques were evaluated by Bioqual veterinary staff before, during and after immunizations. In the macaques studied, complete blood counts and chemistries were obtained throughout the immunization regimen and no marked abnormalities were noted. Of the 10 cynomolgus macaques, there were no adverse events reported at injection sites. Over the course of the study, two cynomolgus macaques experienced slight weight loss. Two cynomolgus macaques showed a single incidence of poor appetite, with one additional cynomolgus macaque showing poor appetite intermittently throughout the study. Additionally, one macaque presented with an infected lymph node biopsy site that responded to appropriate veterinary treatment. Biospecimens were collected before challenge, and 2 and 4 days after challenge, as previously described15. Human samples were obtained with informed consent. All recruitment, sample collection and experimental procedures using human samples have been approved by the Duke Institutional Review Board.

SARS-CoV-2 intranasal and intratracheal challenge

All macaques were challenged at week 11 (3 weeks after last vaccination) through combined intratracheal (3.0 ml) and intranasal (0.5 ml per nostril) inoculation with an infectious dose of 105 plaque-forming units (PFU) of SARS-CoV-2 (2019-nCoV/USA-WA1/2020). The stock was generated at BIOQUAL (lot no. 030120-1030, 3.31 × 105 PFU ml−1) from a p4 seed stock obtained from BEI Resources (NR-52281). The stock underwent deep sequencing to confirm homology with the WA1/2020 isolate. Virus was stored at −80 °C before use, thawed by hand and placed immediately on wet ice. Stock was diluted to 2.5 × 104 PFU ml−1 in PBS and vortexed gently for 5 s before inoculation. Nasal swabs, BAL, plasma and serum samples were collected seven days before, two days after and four days after challenge. Unimmunized control cynomolgus macaques (n = 5) comprised macaques that had been infused with a 10 mg kg−1 of a control antibody (CH65) and then 3 days later challenged with the same challenge dose and stock of SARS-CoV-2 as used in RBD–scNP-immunized macaques or S-2P mRNA-LNP and RBD–scNP-immunized macaques. Protection from SARS-CoV-2 infection was determined by quantitative PCR of SARS-CoV-2 subgenomic E and the more-sensitive N RNA (E or N sgRNA)39 as stated in ‘sgRNA real-time PCR quantification’.

SARS-CoV-2 protein production

The coronavirus ectodomain DNA constructs were synthesized (Genscript), produced and purified as previously described53. S-2P was stabilized by the introduction of 2 prolines at amino acid positions 986 and 987. Plasmids encoding S-2P and HexaPro54 were transiently transfected in FreeStyle 293 cells (Thermo Fisher) using Turbo293 (SpeedBiosystems) or 293Fectin (ThermoFisher). All cells were tested monthly for mycoplasma. The constructs contained an HRV 3C-cleavable C-terminal twinStrepTagII-8×His tag. On day 6, cell-free culture supernatant was generated by centrifugation of the culture and filtering through a 0.8-μm filter. Protein was purified from filtered cell culture supernatants by StrepTactin resin (IBA) and by size-exclusion chromatography using Superose 6 column (GE Healthcare) in 10 mM Tris pH8,150 mM NaCl or 2 mM Tris pH 8, 200 mM NaCl, 0.02% NaN3. ACE2–Fc was expressed by transient transfection of Freestyle 293-F cells53. ACE2–Fc was purified from cell culture supernatant by HiTrap protein A column chromatography and Superdex200 size-exclusion chromatography in 10 mM Tris pH8,150 mM NaCl. SARS-CoV-2 NTD was produced as previously described55. SARS-CoV-2 fusion peptide was synthesized (GenScript).

Sortase A conjugation of SARS-CoV-2 RBD to H. pylori ferritin nanoparticles

Wuhan strain SARS-CoV-2 RBD was expressed with a sortase A donor sequence LPETGG encoded at its C terminus. C-terminal to the sortase A donor sequence was an HRV-3C cleavage site, 8×His tag and a twin StrepTagII (IBA). The SARS-CoV-2 RBD was expressed in Freestyle 293 cells and purified by StrepTactin affinity chromatography (IBA) and Superdex200 size-exclusion chromatography as stated in ‘SARS-CoV-2 protein production’. Helicobacter pylori ferritin particles were expressed with a pentaglycine sortase A acceptor sequence encoded at its N terminus of each subunit. For affinity purification of ferritin particles, 6×His tags were appended C-terminal to a HRV3C cleavage site. Ferritin particles with a sortase A N-terminal tag were buffer exchanged into 50 mM Tris, 150 mM NaCI, 5 mM CaCl2, pH 7.5. Then, 180 μM SARS-CoV-2 RBD was mixed with 120 μM of ferritin subunits and incubated with 100 μM of sortase A overnight at room temperature. Following incubation, conjugated particles were isolated from free ferritin or free RBD by size-exclusion chromatography using a Superose6 16/60 column.

Biolayer interferometry binding assays

Binding was measured using an OctetRed 96 (ForteBio). Anti-human IgG capture (AHC) sensor tips (Forte Bio) were hydrated for at least 10 min in PBS. ACE2 and monoclonal antibodies were diluted to 20 μg ml−1 in PBS and placed in black 96-well assay plate. The influenza antibody CH65 was used as the background reference antibody. The RBD nanoparticle was diluted to 50 μg ml−1 in PBS and added to the assay plate. Sensor tips were loaded with antibody for 120 s. Subsequently, the sensor tips were washed for 60 s in PBS to removed unbound antibody. The sensor tips were incubated in a fresh well of PBS to establish baseline reading before being dipped into RBD–scNP to allow association for 400 s. To measure dissociation of the antibody–RBD–scNP complex, the tip was incubated in PBS for 600 s. At the end of dissociation, the tip was ejected and a new tip was attached to load another antibody. The data were analysed with Data Analysis HT v.12 (ForteBio). Background binding observed with CH65 was subtracted from all values. All binding curves were aligned to the start of association. The binding response at the end of the 400-s association phase was plotted in GraphPad Prism v.9.0.

Surface plasmon resonance assays

Surface plasmon resonance measurements of DH1047 antigen binding fragment (Fab) binding to monomeric SARS-CoV-2 RBD proteins were performed in HBS-EP+ running buffer using a Biacore S200 instrument (Cytiva). Assays were performed in the DHVI BIA Core Facility. The RBD was first captured via its twin-StrepTagII onto a Series S Streptavidin chip to a level of 300–400 resonance units. The antibody Fabs were injected at 0.5 to 500 nM over the captured spike proteins using the single cycle kinetics injection mode at a flow rate of 50 μl min−1. Fab association occurred for 180 s followed by a dissociation of 360 s after the end of the association phase. At the end of the dissociation phase, the RBD was regenerated with a 30-s injection of glycine pH 1.5. Binding values were analysed with Biacore S200 Evaluation software (Cytiva). References included blank streptavidin surface along with blank buffer binding and was subtracted from DH1047 values to account for signal drift and non-specific protein binding. A 1:1 Langmuir model with a local Rmax was used for curve fitting. Binding rates and constants were derived from the curve. Representative results from two independent experiments are shown.

BAL plaque assay

SARS-CoV-2 plaque assays were performed in the Duke Regional Biocontainment biosafery level 3 (BSL3) Laboratory as previously described56. Serial dilutions of BAL fluid were incubated with Vero E6 cells in a standard plaque assay57,58. BAL and cells were incubated at 37 °C and 5% CO2 for 1 h. At the end of the incubation, 1 ml of a viscous overlay (1:1 2× DMEM and 1.2% methylcellulose) was added to each well. Plates are incubated for four days. After fixation, staining and washing, plates were dried and plaques from each dilution of BAL sample were counted. Data are reported as PFU ml−1 of BAL fluid. Samples were collected in virus transport medium from six unimmunized, SARS-CoV-2-challenged macaques for comparison to vaccinated macaques.

SARS-CoV-2 pseudovirus neutralization

For SARS-CoV-2(D614G) and SARS-CoV-2 B.1.1.7 pseudovirus neutralization assays, neutralization of SARS-CoV-2 spike-pseudotyped virus was performed by adapting a previously described infection assay with lentiviral vectors and infection in 293T/ACE2.MF (the cell line was provided by M. Farzan and H. Mu at Scripps). Cells were maintained in DMEM containing 10% FBS and 50 μg ml−1 gentamicin. An expression plasmid encoding codon-optimized full-length spike of the Wuhan-1 strain (VRC7480) was provided by B. Graham and K. Corbett at the Vaccine Research Center, National Institutes of Health. The D614G substitution was introduced into VRC7480 by site-directed mutagenesis using the QuikChange Lightning Site-Directed Mutagenesis Kit from Agilent Technologies (210518). The mutation was confirmed by full-length spike gene sequencing. Pseudovirions were produced in HEK 293T/17 cells (ATCC, CRL-11268) by transfection using Fugene 6 (Promega, E2692). Pseudovirions for infection of 293T cells expressing ACE2 were produced by co-transfection with a lentiviral backbone (pCMV ΔR8.2) and firefly luciferase reporter gene (pHR' CMV Luc)59. Culture supernatants from transfections were clarified of cells by low-speed centrifugation and filtration (0.45-μm filter) and stored in 1 ml aliquots at −80 °C.

For neutralization assays in 293T cells expressing ACE2, a pretitrated dose of virus was incubated with 8 serial threefold or fivefold dilutions of monoclonal antibodies in duplicate in a total volume of 150 μl for 1 h at 37 °C in 96-well flat-bottom poly-l-lysine-coated culture plates (Corning Biocoat). SARS-CoV-2 RBD neutralizing antibody DH1043 spiked into normal human serum was used as a positive control. Cells were suspended using TrypLE express enzyme solution (Thermo Fisher Scientific) and immediately added to all wells (10,000 cells in 100 μl of growth medium per well). One set of eight control wells received cells + virus (virus control) and another set of eight wells received cells only (background control). After 66–72 h of incubation, medium was removed by gentle aspiration and 30 μl of Promega 1× lysis buffer was added to all wells. After a 10-min incubation at room temperature, 100 μl of Bright-Glo luciferase reagent was added to all wells. After 1–2 min, 110 μl of the cell lysate was transferred to a black/white plate (Perkin-Elmer). Luminescence was measured using a PerkinElmer Life Sciences, Model Victor2 luminometer.

To make WA-1, P.1 and B.1.351 SARS-CoV-2 pseudoviruses, human codon-optimized cDNA encoding SARS-CoV-2 spike glycoproteins of various strains were synthesized by GenScript and cloned into eukaryotic cell expression vector pcDNA 3.1 between the BamHI and XhoI sites. Pseudovirions were produced by co-transfection of Lenti‐X 293T cells with psPAX2(gag/pol), pTrip-luc lentiviral vector and pcDNA 3.1 SARS-CoV-2-spike-deltaC19, using Lipofectamine 3000. The supernatants were collected at 48 h after transfection and filtered through 0.45-μm membranes and titrated using HEK293T cells that express ACE2 and TMPRSS2 protein (293-ACE2-TMPRSS2 cells).

For the neutralization assay, 50 μl of SARS-CoV-2 spike pseudovirions were pre-incubated with an equal volume of medium containing serum at varying dilutions at room temperature for 1 h, then virus-antibody mixtures were added to 293T cells expressing ACE2 (WA-1 and B.1.351 assays) or 293-ACE2-TMPRSS2 (WA-1 and P.1 assays) cells in a 96-well plate. After a 3-h incubation, the inoculum was replaced with fresh medium. Cells were lysed 24 h later, and luciferase activity was measured using luciferin. Controls included cell-only control, virus without any antibody control and positive control sera. Neutralization titres are the serum dilution (ID50 or ID80) at which relative luminescence units (RLU) were reduced by 50% or 80%, respectively, compared to virus control wells after subtraction of background RLUs.

Live virus neutralization assays

Full-length SARS-CoV-2, SARS-CoV, WIV-1 and RsSHC014 viruses were designed to express nanoluciferase (nLuc) and were recovered via reverse genetics as previously described60,61,62. Virus titres were measured in Vero E6 USAMRIID cells, as defined by PFU ml−1, in a 6-well plate format in quadruplicate biological replicates for accuracy. For the 96-well neutralization assay, Vero E6 USAMRID cells were plated at 20,000 cells per well the day previously in clear-bottom black-walled plates. Cells were inspected to ensure confluency on the day of assay. Serum samples were tested at a starting dilution of 1:20 and were serially diluted threefold up to nine dilution spots. Serially diluted serum samples were mixed in equal volume with diluted virus. Antibody–virus and virus-only mixtures were then incubated at 37 °C with 5% CO2 for 1 h. Following incubation, serially diluted sera and virus-only controls were added in duplicate to the cells at 75 PFU at 37 °C with 5% CO2. After 24 h, cells were lysed, and luciferase activity was measured via Nano-Glo Luciferase Assay System (Promega) according to the manufacturer specifications. Luminescence was measured by a Spectramax M3 plate reader (Molecular Devices). Virus neutralization titres were defined as the sample dilution at which a 50% reduction in RLU was observed relative to the average of the virus control wells. Prebleed or unimmunized control macaque values were subtracted from WIV-1 neutralization titres, but all other viruses were not background-subtracted.

Biocontainment and biosafety

All work was performed with approved standard operating procedures for SARS-CoV-2 in a BSL3 facility conforming to requirements recommended in the Microbiological and Biomedical Laboratories, by the US Department of Health and Human Service, the US Public Health Service, and the US Center for Disease Control and Prevention and the NIH.

Plasma and mucosal IgG blocking of ACE2 binding

For ACE2 blocking assays, plates were coated with 2 μg ml−1 recombinant ACE2 protein, then washed and blocked with 3% BSA in 1× PBS. While assay plates blocked, purified antibodies were diluted as stated in ‘Plasma and mucosal IgG ELISA binding assays’, only in 1% BSA with 0.05% Tween-20. In a separate dilution plate, S-2P was mixed with the antibodies at a final concentration equal to the half-maximal effective concentration at which spike binds to ACE2 protein. The mixture was allowed to incubate at room temperature for 1 h. Blocked assay plates were then washed and the antibody–spike mixture was added to the assay plates for a period of 1 h at room temperature. Plates were washed and a polyclonal rabbit serum against the same spike protein (S-2P expressed in Freestyle 293-F cells) was added for 1 h, washed and detected with goat anti rabbit-HRP (Abcam, ab97080) followed by TMB substrate. The extent to which antibodies were able to block the binding of the spike protein to ACE2 was determined by comparing the optical density (OD) of antibody samples at 450 nm to the OD of samples containing spike protein only with no antibody. The following formula was used to calculate the percentage of blocking: blocking % = (100 − (OD sample/OD of spike only) × 100).

Plasma and mucosal IgG blocking of RBD monoclonal antibody binding

Blocking assays for DH1041 and DH1047 were performed as stated in ‘Plasma and mucosal IgG blocking of ACE2 binding’, except plates were coated with either DH1041 or DH1047 instead of ACE2.

Plasma and mucosal IgG ELISA binding assays

For ELISA binding assays of coronavirus spike antibodies, the antigen panel included SARS-CoV-2 spike S1 + S2 ECD (SINO, 40589-V08B1), SARS-CoV-2 S-2P53, SARS-CoV-2 spike RBD from mammalian cell 293 (SINO, 40592- V08H), SARS-CoV-2 spike NTD–biotin, SARS-CoV-2 fusion peptide (FP), SARS-CoV spike protein delta (BEI, NR-722), SARS-CoV WH20 spike RBD (SINO, 40150-V08B2), SARS-CoV RBD, MERS-CoV spike S1 + S2 (SINO, 40069-V08B), pangolin GXP4L S-2P, Bat CoV-RaTG13 S-2P, and bat CoV-SHC014 S-2P.

For binding ELISA, 384-well ELISA plates were coated with 2 μg ml−1 of antigens in 0.1 M sodium bicarbonate overnight at 4 °C. Plates were washed with PBS + 0.05% Tween 20 and blocked with assay diluent (PBS containing 4% (w/v) whey protein, 15% normal goat serum, 0.5% Tween-20, and 0.05% sodium azide) at room temperature for 1 h. Plasma or mucosal fluid were serially diluted threefold in superblock starting at a 1:30 dilution. Nasal fluid was started from neat and diluted 1:30, whereas BAL fluid was concentrated tenfold. To concentrate BAL, individual BAL aliquots from the same macaque and same time point were pooled in 3-kDa MWCO ultrafiltration tubes (Sartorious, VS2091). Pooled BAL was concentrated by centrifugation at 3,500 rpm for 30 min or until volume was reduced by a factor of 10. The pool was then aliquoted and frozen at −80 °C until its use in an assay. Purified monoclonal antibody samples were diluted to 100 μg ml−1 and then serially diluted threefold in assay diluent. Samples were added to the antigen-coated plates, and incubated for 1 h, followed by washes with PBS-0.1% Tween 20. HRP-conjugated goat anti-human IgG secondary antibody or mouse anti-rhesus IgG secondary antibody (SouthernBiotech, 2040-05) was diluted to 1:10,000 and incubated at room temperature for 1 h. These plates were washed four times and developed with tetramethylbenzidine substrate (SureBlue Reserve- KPL). The reaction was stopped with 1 M HCl, and OD at 450 nm was determined.

sgRNA real-time PCR quantification

SARS-CoV-2 E gene and N gene subgenomic mRNAs were measured by one-step RT–qPCR adapted from previously described methods63, 64. E and N genes were cloned into pCDNA3.1 and used as in vitro transcription templates. In vitro transcribed RNA was generated with the MEGAscript T7 Transcription Kit (ThermoFisher, AM1334) and purified with MEGAclear Transcription Clean-Up Kit (ThermoFisher, AM1908). Pure RNA was quantified and used as standards for qPCR. RNA extracted from animal samples was quantified using TaqMan Fast Virus 1-Step Master Mix (ThermoFisher, 4444432) and custom primers and probes targeting the E gene sgRNA (forward primer, 5′-CGATCTCTTGTAGATCTGTTCTCE-3′; reverse primer, 5′-ATATTGCAGCAGTACGCACACA-3′; probe, 5′-FAM-ACACTAGCCATCCTTACTGCGCTTCG-BHQ1-3′) or the N gene sgRNA (forward primer, 5′-CGATCTCTTGTAGATCTGTTCTC-3′; reverse primer, 5′-GGTGAACCAAGACGCAGTAT-3′; probe, 5′-FAM-TAACCAGAATGGAGAACGCAGTG GG-BHQ1-3′). A QuantStudio 3 Real-Time PCR System (Applied Biosystems) or a StepOnePlus Real-Time PCR System (Applied Biosystems) was used for real-time PCR reactions. Cycle conditions were as follows: reverse transcription at 50 °C for 5 min, initial denaturation at 95 °C for 20 s, and 40 cycles of denaturation–annealing–extension at 95 °C for 15 s and 60 °C for 30 s. Standard curves were used to calculate E or N sgRNA in copies per ml. The LOD for both E and N sgRNA assays were 12.5 copies per reaction or 150 copies per ml of BAL, nasal swab or nasal wash.

Recombinant IgG production

Expi293-F cells were diluted to 2.5 × 106 cells per ml on the day of transfection. Cells were co-transfected with expifectamine and heavy and light chain expression plasmids. Enhancers were added 16 h after transfection. On day 5, the cell culture was cleared of cells by centrifugation, filtered and incubated with protein A beads overnight. The next day, the protein A resin was washed with Tris buffered saline and then added to a 25-ml column. The resin was washed again and then glacial acetic acid was used to elute antibody off of the protein A resin. The pH of the solution was neutralized with 1 M Tris pH 8. The antibody was buffer-exchanged into 25 mM sodium citrate pH 6 supplemented with 150 mM NaCl, 0.2-μm filtered and frozen at −80 °C.

Negative-stain electron microscopy

The RBD nanoparticle protein at about 1–5 mg ml−1 concentration that had been flash-frozen and stored at −80 °C was thawed in an aluminium block at 37 °C for 5 min; then 1–4 μl of RBD nanoparticle was diluted to a final concentration of 0.1 mg ml−1 into room-temperature buffer containing 150 mM NaCl, 20 mM HEPES pH 7.4, 5% glycerol and 7.5 mM glutaraldehyde. After 5 min of cross-linking, excess glutaraldehyde was quenched by adding sufficient 1 M Tris pH 7.4 stock to give a final concentration of 75 mM Tris and incubated for 5 min. For negative stain, carbon-coated grids (EMS, CF300-cu-UL) were glow-discharged for 20 s at 15 mA, after which a 5-μl drop of quenched sample was incubated on the grid for 10–15 s, blotted and then stained with 2% uranyl formate. After air drying, grids were imaged with a Philips EM420 electron microscope operated at 120 kV, at 82,000× magnification and images captured with a 2k × 2k CCD camera at a pixel size of 4.02 Å.

Processing of negative-stain images

The RELION 3.0 program was used for all negative-stain image processing. Images were imported, CTF-corrected with CTFFIND and particles were picked using a nanoparticle template from previous 2D class averages of nanoparticles alone. Extracted particle stacks were subjected to 2 or 3 rounds of 2D class averaging and selection to discard junk particles and background picks.

Betacoronavirus sequence analysis

Heat maps of amino acid sequence similarity were computed for a representative set of betacoronaviruses using the ComplexHeatmap package in R. In brief, 1,408 betacoronavirus sequences were retrieved from NCBI GenBank, aligned to the Wuhan-1 spike protein sequence and trimmed to the aligned region. The 1,408 spike sequences were then clustered using USEARCH65 with a sequence identity threshold of 0.90, resulting in 52 clusters. We sampled one sequence from each cluster to generate a representative set of sequences. Five betacoronavirus sequences of interest not originally included in the clustered set were added: SARS-CoV-2, GXP4L, bat coronavirus RaTG13, bat coronavirus SHC014 and bat coronavirus WIV-1. This resulted in a set of 57 representative spike sequences. Pairs of spike amino acid sequences were aligned using a global alignment and the BLOSUM62 scoring matrix. For RBD and NTD domain alignments, spike sequences were aligned to the Wuhan-1 spike protein RBD region (residues 330–521) and NTD region (residues 27–292), respectively, and trimmed to the aligned region. Phylogenetic tree construction of RBD sequences was performed with Geneious Prime 2020.1.2 using the neighbour joining method and default parameters. To map group-2b betacoronavirus sequence conservation onto the RBD structure, group-2b spike sequences were retrieved from GenBank and clustered using USEARCH65 with a sequence identity threshold of 0.99, resulting in 39 clusters. For clusters of size >5, 5 spike sequences were randomly downsampled from each cluster. The resulting set of 73 sequences was aligned using MAFFT66. Conservation scores for each position in the multiple sequence alignment were calculated using the trident scoring method67 and computed using the MstatX program (https://github.com/gcollet/MstatX). The conservation scores were then mapped to the RBD domain coordinates (PDB 7LD1) and images rendered with PyMol version 2.3.5.

Histopathology

Lung specimen from macaques were fixed in 10% neutral-buffered formalin, processed and blocked in paraffin for histology analyses. All tissues were sectioned at 5 μm and stained with haematoxylin and eosin to assess histopathology. Stained sections were evaluated by a board-certified veterinary pathologist in a blinded manner. Sections were examined under light microscopy using an Olympus BX51 microscope and photographs were taken using an Olympus DP73 camera.

Immunohistochemistry

Staining for SARS-CoV-2 nucleocapsid antigen was performed by the Bond RX automated system with the Polymer Define Detection System (Leica) following the manufacturer’s protocol. Tissue sections were dewaxed with Bond Dewaxing Solution (Leica) at 72 °C for 30 min, then subsequently rehydrated with graded alcohol washes and 1× Immuno Wash (StatLab). Heat-induced epitope retrieval was performed using Epitope Retrieval Solution 1 (Leica) and by heating the tissue section to 100 °C for 20 min. A peroxide block (Leica) was applied for 5 min to quench endogenous peroxidase activity before applying the SARS-CoV-2 nucleocapsid antibody (1:2,000, GeneTex, GTX135357). Antibodies were diluted in Background Reducing Antibody Diluent (Agilent). The tissue was subsequently incubated with an anti-rabbit HRP polymer (Leica) and colorized with 3,3′-diaminobenzidine chromogen for 10 min. Slides were counterstained with haematoxylin.

Reagent authentication

Cell lines were received with a certificate of authentication certifying their identity. Cell identity was also confirmed by visualizing cell morphology and using flow cytometry to detect cell surface proteins. Cells were confirmed to be free of mycoplasma with monthly testing.

Statistics analysis

Data were plotted using Prism GraphPad 9.0. Wilcoxon rank-sum exact test was performed to compare differences between groups with P value <0.05, considered significant using SAS 9.4 (SAS Institute). No adjustments were made to the P values for multiple comparisons. The 50% and 80% inhibitory dilution (ID50 and ID80, respectively) values were calculated using R statistical software (version 4.0.0). The R package ‘nplr’ was used to fit four-parameter logistic regression curves to the average values from duplicate experiments, and these fits were used to estimate the concentrations corresponding to 50% and 80% neutralization.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.